Human activities drive ecological transformation, impacting island ecosystems from species diversity to ecological traits, mainly through habitat degradation and invasive species. Using two unique long-term datasets we aim to evaluate whether species traits (body size, trophic level, dispersal capacity, and habitat occupancy) can predict temporal variations in the abundance of endemic, Indigenous (endemic and native non-endemic), and exotic arthropods in the Azores Islands. We found that body size is crucial to predict arthropod abundance trends. Small-bodied herbivorous arthropods showed a decrease in abundance, while large-bodied indigenous arthropods increased in abundance, mainly in well-preserved areas. Also, large-bodied exotic arthropods increased in abundance across the entire archipelago. Moreover, endemic canopy dwellers increased in abundance, while endemic ground dwellers decreased in abundance. Simultaneously, exotic arthropods showed the opposite result, increasing abundance in the ground while decreasing abundance in the canopy. Finally, habitat influenced both endemic and exotic spider abundance trends. Endemic spiders that occupy solely natural habitats experienced a decline in abundance, while exotic spiders in the same habitats increased in abundance. Our study underscores the significance of arthropod species traits in predicting abundance changes in island ecosystems over time, as well as the importance of monitoring species communities. Conservation efforts must extend beyond endangered species to protect non-threatened ones, given the increased extinction risk faced by even common species on islands. Monitoring and restoration programs are essential for preserving island ecosystems and safeguarding endemic arthropod populations.

We carried out this study in the Azores Archipelago, which encompasses nine volcanic islands and several islets and seamounts. The archipelago is located in the northern Atlantic Ocean, approximately between 37° and 40° N latitude and 24° and 31° W longitude, spread along a distance of approximately 615 km from East to West. The archipelago is divided into three distinct island groups: western (Corvo and Flores), central (Faial, Pico, São Jorge, Graciosa, and Terceira), and eastern (São Miguel and Santa Maria). On all islands, the climate is characterized as temperate oceanic with high levels of relative atmospheric humidity, which can reach 95% in the elevated native semi-tropical evergreen laurel forest (Elias et al. 2016, Borges et al. 2022a, b, c). Human presence on these islands, dating back to the 15th century, has significantly altered the landscape (Norder et al. 2020). The once-pristine native forests have given way to exotic tree plantations, agricultural and pastoral fields, and urban settlements. Consequently, the original native forest currently occupies a mere 5% of the archipelago's total land area, primarily confined to the remote and elevated regions of difficult access (Gaspar et al. 2008, Triantis et al. 2010, Elias et al. 2016, Borges et al. 2022d).

We sampled arthropods following the standardized sampling protocols of two long-term monitoring projects in the Azores: (i) the ‘Biodiversity of Arthropods from the Laurisilva of the Azores’ (BALA), from 1997 to 2022 (Borges et al. 2006, 2016); and (ii) the ‘Long Term Ecological Study of the Impacts of Climate Change in the Natural Forest of Azores’ (SLAM), from 2012 to 2022 (SLAM), from 2012 to 2022 (Borges et al. 2020, 2022d, Costa and Borges 2021, Lhoumeau and Borges 2023).

We conducted the BALA project in three distinct campaigns, between 1997 and 2004 (BALA I), 2010 and 2011 (BALA II) and 2019 and 2022 (BALA III). We collected arthropods from native forest fragments on seven islands. Corvo and Graciosa were not sampled due to the lack of native forest patches. Two sampling methods were used: pitfall traps to capture ground-dwelling arthropods and canopy beating to capture canopy-dwelling arthropods (maximum height of 3-4 m). All the samplings occurred in boreal summer, between June and September. More details of the methods can be found elsewhere (Ribeiro et al. 2005, Borges et al. 2006, 2016, Gaspar et al. 2008) and data is now openly available (Pozsgai et al. 2024).

We collected the SLAM project data once in each boreal season (spanning three months), continuously between 2012 and 2022. Sampling was conducted in a total of 10 plots, located only on Terceira island. Plots were set in areas that are among the best-preserved native forests fragment, with minimal human disturbance. Arthropods were sampled using passive flight interception SLAM traps (Sea, Land and Air Malaise trap) at each plot. Each trap had dimensions of 110 cm x 110 cm x 110 cm and contained propylene glycol (pure 1,2-propanediol) to capture, kill and conserve the specimens (Borges et al. 2020, 2022d, Costa and Borges 2021, Lhoumeau and Borges 2023). More details of the methods can be found elsewhere (Borges et al. 2020, 2022d, Costa and Borges 2021, Lhoumeau and Borges 2023).

Most of the collected arthropods were sorted to species level, and nomenclature follows the most recent checklist of Azorean arthropods (Borges et al., 2022a). The remaining specimens that were not identified to species level were sorted to morphospecies and were excluded from the analyses. Also, all Crustacea, Collembola, Diplura, Diptera and Hymenoptera (except Formicidae) species were not considered in this study. Nominated species were then classified according to their biogeographic origin, into the following three categories (Borges et al. 2022a): i) endemic species, i.e. those exclusive to the Azores; ii) indigenous species, i.e. those that are a combination of endemic and native non-endemic Azorean species, which can also be found in neighboring archipelagos (Madeira and the Canary Islands) and/or in the Mediterranean basin, and most likely reached the Azores through long-distance dispersal; and iii) exotic species, i.e. those believed to have reached the Azores by human actions (Borges et al., 2006, 2022a; Borges and Wunderlich, 2008).

For all sampled species, we selected a total of six functional traits for which we had a priori expectations about how they may respond to increasing habitat disturbance (e.g. body size, dispersal ability), traits that are important in species interaction (e.g. trophic status, specialization) and traits that are common to all arthropod groups encountered in our study (Gossner et al. 2015, Simons et al. 2016, Rigal et al. 2018, Chichorro et al. 2022a) (Fig. 1). The six traits were: (i) body size, that was standardized by the z-score within each order, i.e., average size of the order minus the size of the species divided by the standard deviation of size in the order, (ii) vertical stratum that the species occupy in the vegetation (henceforth verticality), expressed as average verticality (AVS), from 0 to 1 where 0 is exclusive ground dwelling species and 1 is the exclusive canopy dwelling species, (iii) feeding trophic group, expressed in four categories (carnivorous, fungivorous, herbivorous and omnivorous); dispersal ability within and between islands, classified into (iv) four categories (ballooning, crawling, flying and phoretic) and into (v) low (crawl, phoretic) or high (ballooning and flying) dispersal capacities; and (vi) habitat occurrence (where species were present) as categorical variables (presence in natural habitat, in anthropogenic habitat or in both natural and anthropogenic habitat) (see more details in Table 1). In the case of the verticality, we followed the framework described in Costa et al. (2023). Using the BALA databases (I to III), we attributed a verticality score to the sample types, with 0 assigned to pitfalls, 0.5 to beating samples from bushes, and 1 to trees. The relative abundance for each species across each stratum in the forest fragments was obtained dividing its abundance by the total number of spiders sampled in that stratum and site was standardized for variations in the sampling method. Then, we normalized the values by dividing each species' stratum relative abundance by the sum of its values along all strata within each forest fragment. All these were calculated using the equations described for average verticality in Costa et al. (2023). When we analyzed spiders separately, only three traits were used: (i) body size, (ii) average verticality (AVS) and (iii) habitat occurrence. Due to the homogeneity of spiders’ feeding trophic level (all carnivores) and dispersal ability (only one species was a crawling spider, all others were considered possible ballooners), we removed these traits from the analysis. Moreover, habitat occurrence for spiders only had two levels and were analyzed as a binary variable (inhabiting only natural habitats or both natural and anthropogenic habitats). Apart from body size, measured on the individuals sampled in these studies, data for the functional traits were compiled from: (i) the Azorean Biodiversity Portal (ABP 2024) or IUCN SSC Atlantic Islands Invertebrate Group Portal (ABP 2024, AIISG 2024); (iii) published data, papers and descriptions of particular species; and (iv) personal knowledge on the species’ natural history by PAVB.

We partitioned the data for the analyses into six different datasets, of which three were the Scenario Analysis and three were the Sensitivity Analysis (Supplementary Material). For the Scenario Analysis, we used the complete BALA dataset (hereafter BALA full) and the SLAM dataset. Since the SLAM data is limited to Terceira island, we also used the BALA data from Terceira only (hereafter BALA Terceira) to compare the results obtained from the two sampling strategies (Fig. 1 and Supplementary Table 1). For the Sensitivity Analysis, we used the BALA dataset excluding Pico island (hereafter BALA without Pico), the BALA dataset excluding Terceira island (hereafter BALA without Terceira), and the BALA dataset excluding Pico and Terceira islands (hereafter BALA without Pico and Terceira). Since the native forest is not equally preserved across all Azores islands, we tested the removal of Pico and Terceira islands, which were the most preserved native forests throughout the archipelago (Tsafack et al. 2023) (Supplementary Table 1).

Based on the time length of our datasets (25 years for BALA and 10 years for SLAM) and to improve the reliability of trends and traits analyses, we excluded species with less than 10 adult individuals in each of BALA full or SLAM datasets. In this way, we mitigate the potential data noise caused by rare species (Martínez-Núñez et al. 2024), which we considered as those with one or less individuals per year in ten years or more. Then, we compiled the abundance of each species per: (i) BALA sampling (BALA I, II and III) and island sites (30), for all BALA datasets and (ii) SLAM years (10 years) and sites (10) for the SLAM dataset (Supplementary Table 1). In this way, we were able to obtain a abundance trend (slope) of each species in each dataset (see below). Each dataset was further separated into four groups of species: all arthropods, only herbivorous arthropods, only spiders and only beetles, the last two being the most diverse groups of the sampled arthropods. Also, the same groups were studied according to their biogeographic origins: all species together, only endemic species, only indigenous species and only exotic species (Supplementary Table 1). Therefore, we had a total of 16 separate analyses for each dataset (four arthropod groups by four biogeographic origins; 64 Bayesian frameworks, see below).

To obtain the abundance trend of each species in each dataset (slope values, β), we fitted null (intercept only) GLMM models with Negative Binomial distribution and random effect (Goldstein and de Valpine 2022). The response variables were each species abundance (one species per model) and the random effects were specific to each dataset: for BALA full, BALA without Pico, BALA without Terceira, and BALA without Pico and Terceira, islands were the random variable (seven islands); and for BALA Terceira and SLAM, sites were the random variables (10 sites). Hence, the intercept of the response variable with a positive estimate indicated a positive abundance trend over time, while a negative estimate value indicated a negative abundance trend over time. GLMM models were performed and Conditional R-squared (delta method, to estimate standard errors of transformations of a random variable, a first-order Taylor approximation (Parr 1983)) were obtained through the MuMIn package (Bartoń 2023) in the R environment (R Team 2022).

We fitted a Bayesian framework, weighted by Conditional R-squared values obtained for each species, to assess how changes in species abundance trends (response variable) could be predicted by species traits (explanatory variables) (Fig. 1). Distinct random effects were chosen to fit each framework for each dataset, being either the arthropod order and/or the species origin (Supplementary Table 1). For instance, when we analyzed all arthropods, both order (each arthropod order) and origin (endemic, native non-endemic or exotic) were used as random effects. At the same time, when indigenous arthropods were analyzed, order was used as previously (each arthropod order) but origin was only endemic or native non-endemic, as random effects (Supplementary Table 1). All Bayesian frameworks were built with the help of the jagsUI package (Kellner and Meredith 2021) in R environment (R Team 2022), with five simultaneous chains, default priors, one million interactions, 800.000 burn-ins and 10 thins. Variables were considered to be significant when 97.5% of the credible intervals (CRI) were either positive or negative, not passing through zero (Hespanhol et al. 2019).

References

ABP 2024. Azorean Biodiversity Portal.

AIISG 2024. Atlantic Island Invertebrates Specialist Group.

Bartoń, K. 2023. Mu-MIn: Multi-model inference. in press.

Borges, P. A. V. and Wunderlich, J. 2008. Spider biodiversity patterns and their conservation in the Azorean archipelago, with descriptions of new species. - Syst. Biodivers. 6: 249–282.

Borges, P. A. V., Lobo, J. M., De Azevedo, E. B., Gaspar, C. S., Melo, C. and Nunes, L. V. 2006. Invasibility and species richness of island endemic arthropods: A general model of endemic vs. exotic species. - J. Biogeogr. 33: 169–187.

Borges, P. A. V., Gaspar, C., Crespo, L. C. F., Rigal, F., Cardoso, P., Pereira, F., Rego, C., Amorim, I. R., Melo, C., Aguiar, C., André, G., Mendonça, E. P., Ribeiro, S., Hortal, J., Santos, A. M. C., Barcelos, L., Enghoff, H., Mahnert, V., Pita, M. T., Ribes, J., Baz, A., Sousa, A. B., Vieira, V., Wunderlich, J., Parmakelis, A., Whittaker, R. J., Quartau, J. A., Serrano, A. R. M. and Triantis, K. A. 2016. New records and detailed distribution and abundance of selected arthropod species collected between 1999 and 2011 in Azorean native forests. - Biodivers. Data J. 4: 1–84.

Borges, P. A. V., Rigal, F., Ros-Prieto, A. and Cardoso, P. 2020. Increase of insular exotic arthropod diversity is a fundamental dimension of the current biodiversity crisis. - Insect Conserv. Divers. 13: 508–518.

Borges, P. A. V., Lamelas-Lopez, L., Andrade, R., Lhoumeau, S., Vieira, V., Soares, A. O., Borges, I., Boieiro, M., Cardoso, P., Crespo, L. C. F., Karsholt, O., Schülke, M., Serrano, A. R. M., Quartau, J. A. and Assing, V. 2022a. An updated checklist of Azorean arthropods (Arthropoda). - Biodivers. Data J. in press.

Borges, P. A. V., Lamelas-López, L., Tsafack, N., Boieiro, M., Ros-Prieto, A., Gabriel, R., Nunes, R. and Ferreira, M. T. 2022b. SLAM Project - Long Term Ecological Study of the Impacts of Climate Change in the Natural Forest of Azores: III - Testing the impact of edge effects in a native forest of Terceira Island. - Biodivers. Data J. in press.

Borges, P. A. V., Lamelas-Lopez, L. and Schülke, M. 2022c. New records of rove-beetles (Insecta, Coleoptera, Staphylinidae) for Azores Islands (Portugal). - Biodivers. Data J. in press.

Borges, P. A. V., Lamelas-Lopez, L., Stüben, P. E., Ros-Prieto, A., Gabriel, R., Boieiro, M., Tsafack, N. and Ferreira, M. T. 2022d. SLAM Project - Long Term Ecological Study of the Impacts of Climate Change in the Natural Forest of Azores: II - A survey of exotic arthropods in disturbed forest habitats. - Biodivers. Data J. in press.

Campomizzi, A. J., Butcher, J. A., Farrell, S. L., Snelgrove, A. G., Collier, B. A., Gutzwiller, K. J., Morrison, M. L. and Wilkins, R. N. 2008. Conspecific Attraction is a Missing Component in Wildlife Habitat Modeling. - J. Wildl. Manage. 72: 331–336.

Chichorro, F., Urbano, F., Teixeira, D., Väre, H., Pinto, T., Brummitt, N., He, X., Hochkirch, A., Hyvönen, J., Kaila, L., Juslén, A. and Cardoso, P. 2022a. Trait-based prediction of extinction risk across terrestrial taxa. - Biol. Conserv. 274: 0–7.

Chichorro, F., Correia, L. and Cardoso, P. 2022b. Biological traits interact with human threats to drive extinctions: A modelling study. - Ecol. Inform. 69: 101604.

Costa, R. and Borges, P. A. V. 2021. SLAM Project - Long Term Ecological Study of the Impacts of Climate Change in the natural forest of Azores: I - the spiders from native forests of Terceira and Pico Islands (2012-2019). - Biodivers. Data J. 9: 1–28.

Costa, R., Cardoso, P., Rigal, F. and Borges, P. A. V. 2023. Island spider origins show complex vertical stratification patterns in Macaronesia. - Insect Conserv. Divers.: 1–10.

Elias, R. B., Gil, A., Silva, L., Fernández-Palacios, J. M., Azevedo, E. B. and Reis, F. 2016. Natural zonal vegetation of the Azores Islands: Characterization and potential distribution. - Phytocoenologia 46: 107–123.

Gaona, F. P., Iñiguez-Armijos, C., Brehm, G., Fiedler, K. and Espinosa, C. I. 2021. Drastic loss of insects (Lepidoptera: Geometridae) in urban landscapes in a tropical biodiversity hotspot. - J. Insect Conserv. 25: 395–405.

Gaspar, C., Borges, P. A. V and Gaston, K. J. 2008. Diversity and distribution of arthropods in native forests of the Azores archipelago. - Arquipélago 25: 1–30.

Goldstein, B. R. and de Valpine, P. 2022. Comparing N-mixture models and GLMMs for relative abundance estimation in a citizen science dataset. - Sci. Rep. 12: 1–12.

Gossner, M. M., Simons, N. K., Achtziger, R., Blick, T., Dorow, W. H. O., Dziock, F., Köhler, F., Rabitsch, W. and Weisser, W. W. 2015. A summary of eight traits of Coleoptera, Hemiptera, Orthoptera and Araneae, occurring in grasslands in Germany. - Sci. Data 2: 1–9.

Hespanhol, L., Vallio, C. S., Costa, L. M. and Saragiotto, B. T. 2019. Understanding and interpreting confidence and credible intervals around effect estimates. - Brazilian J. Phys. Ther. 23: 290–301.

Kehoe, R., Frago, E. and Sanders, D. 2020. Cascading extinctions as a hidden driver of insect decline. - Ecol. Entomol. in press.

Kellner, K. and Meredith, M. 2021. jagsUI - A Wrapper Around “rjags” to Streamline “JAGS” Analyses.: 1–18.

Lhoumeau, S. and Borges, P. A. V. 2023. Assessing the Impact of Insect Decline in Islands: Exploring the Diversity and Community Patterns of Indigenous and Non-Indigenous Arthropods in the Azores Native Forest over 10 Years. - Diversity 15: 753.

Martínez-Núñez, C., Gossner, M. M., Maurer, C., Neff, F., Obrist, M. K., Moretti, M., Bollmann, K., Herzog, F., Knop, E., Luka, H., Cahenzli, F. and Albrecht, M. 2024. Land-use change in the past 40 years explains shifts in arthropod community traits. - J. Anim. Ecol.: 1–14.

Mawan, A., Hartke, T. R., Deharveng, L., Zhang, F., Buchori, D., Scheu, S. and Drescher, J. 2022. Response of arboreal Collembola communities to the conversion of lowland rainforest into rubber and oil palm plantations. - BMC Ecol. Evol. 22: 1–13.

Munstermann, M. J., Heim, N. A., McCauley, D. J., Payne, J. L., Upham, N. S., Wang, S. C. and Knope, M. L. 2021. A global ecological signal of extinction risk in terrestrial vertebrates. - Conserv. Biol. 36: 1–13.

Norder, S. J., de Lima, R. F., de Nascimento, L., Lim, J. Y., Fernández-Palacios, J. M., Romeiras, M. M., Elias, R. B., Cabezas, F. J., Catarino, L., Ceríaco, L. M. P., Castilla-Beltrán, A., Gabriel, R., de Sequeira, M. M., Rijsdijk, K. F., Nogué, S., Kissling, W. D., van Loon, E. E., Hall, M., Matos, M. and Borges, P. A. V. 2020. Global change in microcosms: Environmental and societal predictors of land cover change on the Atlantic Ocean Islands. - Anthropocene in press.

Parr, W. C. 1983. A note on the jackknife, the bootstrap and the delta method estimators of bias and variance. - Biometrika 70: 719–722.

Pozsgai, G., Lhoumeau, S., Amorim, I. R., Boieiro, M., Cardoso, P., Costa, R., Ferreira, M. T., Leite, A., Malumbres-olarte, J., Oyarzabal, G., Rigal, F., Ros-prieto, A., Santos, A. M. C., Gabriel, R. and Borges, P. A. V 2024. The BALA project : A pioneering monitoring of Azorean forest invertebrates over two decades.: 1–9.

Purvis, A., Gittleman, J. L., Cowlishaw, G. and Mace, G. M. 2000. Predicting extinction risk in declining species. - Proc. R. Soc. B Biol. Sci. 267: 1947–1952.

R Core Team 2023. R : A Language and Environment for Statistical Computing. - R Found. Stat. Comput. in press.

Reuter, D. M., Hopkins, S. S. B. and Price, S. A. 2023. What is a mammalian omnivore? Insights into terrestrial mammalian diet diversity, body mass and evolution. - Proc. R. Soc. B Biol. Sci. in press.

Ribeiro, S. P., Borges, P. A. V., Gaspar, C., Melo, C., Serrano, A. R. M., Amaral, J., Aguiar, C., André, G. and Quartau, J. A. 2005. Canopy insect herbivores in the Azorean Laurisilva forests: Key host plant species in a highly generalist insect community. - Ecography (Cop.). 28: 315–330.

Rigal, F., Cardoso, P., Lobo, J. M., Triantis, K. A., Whittaker, R. J., Amorim, I. R. and Borges, P. A. V. 2018. Functional traits of indigenous and exotic ground-dwelling arthropods show contrasting responses to land-use change in an oceanic island, Terceira, Azores. - Divers. Distrib. 24: 36–47.

Rocha-Ortega, M., Rodríguez, P. and Córdoba-Aguilar, A. 2019. Spatial and temporal effects of land use change as potential drivers of odonate community composition but not species richness. - Biodivers. Conserv. 28: 451–466.

Simons, N. K., Weisser, W. W. and Gossner, M. M. 2016. Multi-taxa approach shows consistent shifts in arthropod functional traits along grassland land-use intensity gradient. - Ecology 97: 754–764.

Triantis, K. A., Borges, P. A. V., Ladle, R. J., Hortal, J., Cardoso, P., Gaspar, C., Dinis, F., Mendonça, E., Silveira, L. M. A., Gabriel, R., Melo, C., Santos, A. M. C., Amorim, I. R., Ribeiro, S. P., Serrano, A. R. M., Quartau, J. A. and Whittaker, R. J. 2010. Extinction debt on oceanic Islands. - Ecography (Cop.). 33: 285–294.

Tsafack, N., Pozsgai, G., Boieiro, M., Ros-Prieto, A., Nunes, R., Ferreira, M. T. and Borges, P. A. V. 2023. Edge effects constraint endemic but not introduced arthropod species in a pristine forest on Terceira (Azores, Portugal). - For. Ecol. Manage. in press.

Van den Broeck, M., Rhazi, L., Waterkeyn, A., El Madihi, M., Grillas, P., Kneitel, J. M. and Brendonck, L. 2019. Livestock disturbances in Mediterranean temporary ponds: A mesocosm experiment with sheep manure and simulated trampling. - Freshw. Biol. 64: 856–869.

Zhang, J., Saqib, H. S. A., Niu, D., Guaman, K. G. G., Wang, A., Yu, D., You, M., Pozsgai, G. and You, S. 2023. Contrasting roles of landscape compositions on shaping functional traits of arthropod community in subtropical vegetable fields. - Agric. Ecosyst. Environ. 347: 108386.